RESEARCH ACCOUNT Relative roles of computational fluid dynamics and wind tunnel testing in the development of

S. S. Desai Computational and Theoretical Fluid Dynamics Division, National Aerospace Laboratories, Wind Tunnel Centre, Belur Campus, Bangalore 560 037, India

may make one believe that the wind tunnels can hence- Much is talked of Computational Fluid Dynamics forth be laid to rest in our efforts at designing and deve- (CFD), concerning its role in the possibility of bring- loping an aircraft. At the present time it is, for instance, ing down the costs and turn-around times in the design and development of aircraft. Views have been natural for anyone to ask the question: is it really true? expressed, at one extreme end, that wind tunnel test- Or else, how much of wind tunnel testing have we saved ing may play a secondary role eventually. This com- as a result of the CFD efforts? These questions are rele- munication takes a balanced view according to which vant because of the tremendous support that CFD has both CFD and wind tunnels are inevitable; there are enjoyed over the past two to three decades, with a pro- roles which are exclusive for CFD and wind tunnels mise of providing a new tool to the designer. and there are roles which are synergistic and com- The answer to this question is not simple. While some plementary. Additionally, both CFD and wind tunnel enthusiasts of CFD might like to insist that sooner or testing are, after all, only tools whose optimality of later wind tunnel testing may have to play only second usage is always as good as the experience and the fiddle in the aircraft development, practical reality under- relevant design database of the designer of the air- scores the importance this wind tunnel testing has been craft. It is argued here that the efficacy of CFD in aircraft playing, and is certain to continue to do so, in the aero- design is to be judged not so much by the number of nautical development activities; there are indeed areas of wind tunnel blow downs that can be saved as it is to fluid flow which can only be handled effectively by wind be judged by the value addition to the design cycle by tunnel testing. In this exclusive domain it is unlikely that its contribution to a superior baseline for final tuning our theoretical understanding of fluid flows, or of model- and check-out of configuration through detailed wind ling them mathematically or their numerically simulation tunnel testing and by the guidance obtainable through would ever ripen to a stage where wind tunnel testing rapid CFD simulations for better utilization of tunnel will become unnecessary. At the same time, there are tests. also areas which are exclusive for CFD. For example, CFD is extremely well suited in the evolution of configu- PRIOR to the mid-seventies, aircraft design was mainly rations, particularly for optimized performance at a single based on approximate theories of fluid flow, on engineer- design point, such as, for example, the cruise perform- ing data sheets and relied heavily on vast amount of wind ance for a transport aircraft; the flow domains under such tunnel testing. Even then human ingenuity ensured conditions are predominantly well behaved with domi- superb designs such as the Concords and the Jumbos. nantly attached flows which imply small boundary layer Over the last two decades, however, the discipline of effects. This is an area where CFD finds itself at ease. Computational Fluid Dynamics (CFD) is fast changing There are also areas where CFD simulations can be this world of airplane design1,2. A new tool is now avail- synergistically used with wind tunnel testing; with the able to the designer in the form of a variety of CFD soft- help of detailed flow simulation obtainable by a CFD ware with varying degree of sophistication and accuracy simulation, it is possible to organize flow diagnostics and of flow simulation. metric elements optimally in any wind tunnel testing. The reputation which this vastly developed discipline Finally, the role of experience and support from design has gained over the past decade tends to lend support to data bank in the development of superior aircraft an utopian view that aircraft can henceforth be designed configurations cannot be ignored. It is indeed the experi- by the computer keyboard alone. The much-advertised ence which actually ensures an optimum utilization of the statements in open literature about the CFD capabilities, essential design clues, not only from CFD and wind tunnel tests but also from approximate theories1, in order to bring about competitive designs within reduced e-mail: [email protected] cycle times and with cost benefits. Subtract this experi-

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 49 RESEARCH ACCOUNT ence, and any of these sophisticated tools become in- The edge of CFD simulations over wind tunnel tests effective. lies in their ability to assess swiftly different configura- tions at one or two design points for better performance. In the pre-CFD era this step was taken by the designer Essential elements in the aerodynamic develop- through approximate theories, available proprietary data ment of aerospace configurations and through design data sheets as shown in Figure 2. In this mode, wind tunnel tests were not only used to vali- There are many areas of activity involved in the design date the candidate design but also, in a limited way, used and development of aerospace configurations. These in the preliminary design stage of an aircraft. Refinement include: , structural design and analysis, to a candidate configuration in the preliminary design power plant and its integration with aircraft, system inte- stage through wind tunnel testing could be prohibitively gration and flight testing. In this article, we concentrate expensive due to the model making costs and tunnel run- only on the activities related to aerodynamic configura- ning costs. Because of both time and cost elements tion design, which is the first and the foremost step in involved in wind tunnel testing, it becomes impractical to any aircraft design. study several major configurational changes during the Figure 1 depicts the four elements that are essential in initial design freeze of the shapes. This could have pre- the design and development of any aircraft, or for that vented a superior candidate being frozen at the end of the matter any aerospace configuration such as missiles, preliminary design phase. launch vehicles, etc. The laudable progress made in the ability of CFD It is to be noted here that it is the experience tempered simulations for complex 3D, real-life configurations now by database, wind tunnel testing and CFD simulations gives CFD the prime role in the fast design freeze of the that is at the root of any aircraft design. Experience is of candidate configuration (Figure 3). considerable significance in de novo designs, and it It is to the credit of CFD developments that, for exam- dominates, particularly in all the derivative designs ple, aerofoil analysis and design and optimization of aimed at improving a given design, either to enhance its wing–body combinations can now be carried out with a performance or to suit it best to the changing market/ high level of confidence through CFD. One can now consumer requirements. It is once again the experience at design an aerofoil for laminar or supercritical applica- Boeing that made it possible for CFD simulations to play a major role in the final design decisions in the develop- ment of the 777 airplane, pushing the role of wind tun- nels into the ‘after-the-fact’ design validation2. Figure 1 also shows, by dashed lines, contributions to wind tunnel testing and to data sheets that are becoming possible in recent times owing to latest progress made in more realistic flow simulation through CFD. These were not available some 10–15 years ago.

Exclusive roles of CFD and wind tunnel testing Limited Wind There are appropriate roles of wind tunnel tests and CFD simulations that are exclusive to each activity. The wind tunnels, like the flight tests, dominate the phase of design fine-tuning and design validation.

Figure 2. Typical aircraft design process during pre-CFD era. Note the essential role of the wind tunnels in the preliminary design phase of an aircraft. This role, essential in the pre-CFD era, is now taken over Figure 1. Elements in the design of aerodynamic shapes. by CFD (see Figure 3).

50 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT tions instead of choosing them from NACA or other design conditions which are primarily dominated by experimental databases which were the only source dur- medium or large-scale flow separations, flight Reynolds ing pre-CFD era. Similarly, by an inverse design it is number effects, control-surface gap effects, control- possible to arrive at, quickly during a couple of CFD surface effectiveness, installed power-system losses, simulations, an optimized shape to give some advanta- combined internal and external flows and so on. Figure 3 geous pressure distribution which is conducive to mini- shows the present-day scenario in which CFD now mizing shock strength or avoiding originating at the occupies an important step in the initial design stage. wing tips or constraining the behaviour of pitching Here, not only can CFD reduce the cycle time in the final moment, etc. design cycle by minimizing the possibility of the The tunnel experiments, on the other hand, have a changes to baseline, which are typically brought to the pivotal role to play in the validation of CFD software. surface during proving flight tests, but it can also give the For this purpose carefully planned and specific experi- designer an advantage in cost and cycle times through the ments would be required. There is still a pressing need input of a superior baseline to begin the essential design for basic experiments to provide validational data for process. In fact, it is in this manner that CFD has its CFD in the area of flow stability, transitional flows, hys- major contribution. terisis effects and the vast area of and its Figure 4, adapted from Rubbert2 is more instructive modelling, and for three-dimensional boundary layers in bringing out the relative roles of CFD and wind and flow separation characteristics. tunnels. What is significant is the fact that in a typical The wind tunnels hold and continue to hold their aircraft design, CFD may be able to contribute through supremacy in the design validation of configurations at mere hundreds of simulations; beyond this the cost and the off-design conditions. The major issues that do not time for simulations would be prohibitive. The wind easily lend themselves to CFD analysis include: off- tunnel on the other hand, has to cater to the requirement of millions of aerodynamics simulations that are invaria- bly essential to validate any design before fabrication of the prototypes. It is to be noted here that one simulation consists of one free stream and one incidence for one geometry configuration. Figure 4, a typical experience of aircraft companies, brings out the essential role of CFD, more importantly in the initial design phase through a small number of flow simulations where it is cost and time-effective, whereas the wind tunnel is expensive in an iterative process for the choice of the baseline configurations. After the initial time and cost of wind tunnel model fabrication, the wind tunnel is more expedient for the large number of simulations typically required in any aircraft design.

Figure 4. Cost and flow time characteristics of wind tunnels and CFD2. It is noteworthy that the cost/time of CFD simulations are not Figure 3. Changing world of airplane design with the emergence of affordable beyond 10 s and 100 s of simulations. With the availability CFD as an important design tool, particularly in the preliminary design of a model and tunnel time slot, wind tunnel simulations are much phase. cheaper.

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 51 RESEARCH ACCOUNT

Areas of uncertainly in flow simulation which are Advantages in CFD simulations common to both CFD and wind tunnels The CFD techniques are particularly effective in the In a sense, neither CFD nor wind tunnels are perfect following efforts: as far as simulation of realistic flow fields is concerned. This is understandable. Not only does each one of them i. Aerofoil design and analysis, and optimization of have its own drawbacks, there are also areas common to wing–body shapes. both CFD and wind tunnels where there are questions of ii. Deciphering in a short time the merits and demerits difficulties and of their accuracies. These areas are: of several variants considered for a baseline. A few · High flows – Simulation of full-scale simulations at one design condition can reveal Reynolds number (Re) of the order of 20–40 million is merits of one variant over the other. unusual and out of the question for wind tunnels for iii. Detailed flow field information. There are times reasons like cost of building and operation of the where flow field information in specific locations facility. CFD simulation has also a lot of uncertainties over the configuration may be required (For example, here because of grid and turbulence model limitations. the flow distribution at the inlet plane of an engine.) · Drag evaluation – Drag evaluation within 5 to 10% This is not only too cumbersome in wind tunnels but accuracy is what is typically possible in both CFD and also, often impractical. wind tunnel simulations. iv. Quicker and less expensive changes to configuration · Model support and wall interference in wind tunnels and their numerical simulation. and influence of outer boundaries in CFD – These are v. Smoothing of the configuration to reduce pressure- elements which contribute to interference effects. drag levels. · Flow at large incidence; CLmax for a wing, for example – vi. Quick checkouts during after-design stages, where it Unduly large blockage, and tunnel-wall effects is not possible to carry out wind tunnel tests for combined with scale and free stream turbulence effects want of timely tunnel slots or logistics. result in uncertainties in tunnel measurements. In CFD, vii. Assessment of incremental changes quickly in the modelling of large-scale separation is still in a nascent configuration. state; lack of turbulence models is a big hindrance. viii. Design of air-data system and determination of · Gap effects of control surfaces – Re (scale) effects in suitable location for air-data probes on a given gaps can give misleading results in wind tunnel testing. aircraft. Also, model inaccuracy is a contributing factor in the ix. Hypersonic flow and chemical/rarefied-gas effects, small sizes of models that are practical for tunnel testing. for which experiments are impossible to be done. · Confluent boundary layers and internal–external flows – Close geometric simulation of internal and external flows is a herculean task in tunnel testing. Advantages of wind-tunnel testing · Resolution of viscous effects – Effects of free stream i. Accurate data generation for validating CFD turbulence in tunnels and of turbulence models in CFD simulations. CFD has considerably benefited simulations contribute to simulation errors. from this role of wind tunnel simulation. · Power effects – This poses considerable complications ii. Simulation of complex shapes where grid gen- in tunnel testing and is not amenable to CFD. eration and accuracy of flow simulation pose · Implementational difficulties – The elements of cost definitive problems in CFD. and time of model making and building a tunnel facil- iii. Simulation of separated flows, typical of large ity, its maintenance and difficulty of availability of incidence aerodynamics. Here limitations of trained manpower to man the tunnels (in this domi- accuracy of turbulence models is a big hindrance nantly Information Technology era) are a big deterrent in CFD simulations. This area would be domi- to tunnel testing. Similarly, the CFD software is hin- nated by wind tunnels for a long time to come. dered by the necessity of long hours of set-up time and iv. Simulation of viscous effects in corners and over of longer apprenticeship needed to develop adequate gaps in control surfaces. familiarity with limitations, range of applicability and v. Expediency in store carriage and separation with capabilities of a CFD software. studies. vi. Simulation of power effects, rotor flows and Areas of individual superiority for CFD and multi-component flows with relative ease. wind tunnels vii. In one tunnel-run lasting for less than a minute it is possible to obtain 15–20 simulations, whereas There are areas in flow simulation where wind tunnels in CFD simulations one run of the code gener- and CFD have their own advantages. ates only one simulation.

52 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT

Questions of CFD code validation of the CFD codes, including the most sophisticated ones. However, overall characteristics may be reliably This is a truly difficult issue. It is to be constantly kept in obtained from many of the CFD codes. Similarly with mind that CFD tools cannot be used as a black-box by some care overall pitching moment behaviour – for anyone without much exposure to the nature of CFD, its example, stable or unstable behaviour – can be surmised capabilities, its weakness and pitfalls that are likely to by a few other CFD codes12,13 (Figure 5). Many RANS arise in any blind use. codes, for instance, can predict the quantitative pitching CFD is presently a fairly mature tool for the analysis moments and hence also the detailed pressure distribu- and design of 2D shapes and in particular aerofoils. A tions for two-dimensional flows reasonably well3,9, par- variety of software can be used for this purpose: tran- ticularly when the configuration and the flow are not sonic full-potential codes with boundary-layer correction, prone to large viscous effects or flow separations. for both analysis3 and design of supercritical aerofoils4, A hierarchy of flow simulation codes is therefore simple codes for design of laminar-flow profiles5,6, essential for an effective functional use of CFD in any sophisticated Re-averaged Navier–Stokes codes7,8 for vehicle development work. This is not in any way dif- analysis of viscous flows even close to stall9,10 as well as ferent from what is true with wind tunnel testing; special- unsteady flows11. CFD tools can also be used with fair ized tunnels are indeed necessary for specific tasks: water amount of confidence in the analysis of wing–body tunnels for flow visualization, particularly for - shapes or in their optimization. Simple wing-alone cases dominated flows; specialized quiet tunnels for laminar- are amenable to reasonably good CFD simulations. flow investigations; special low-speed tunnels for high- Extremely complex 3D, real-life shapes continue to lift investigations; 2D-tunnels for aerofoil pose, of course, considerable challenge to CFD tools. So analysis; large low-speed tunnels for power-effects and are also flow regimes dominated by separated viscous full high-lift configuration, special tunnels for rotary flows, confluent viscous flows over multiple components, flows and so on. etc. Even then, it is possible to grossly assess the What is to be specially emphasized here is the follow- configuration. Here the panel codes12, despite being only ing observation: CFD is a tool and the efficacy of its use limited to inviscid flow simulations, have been the work- depends on the user; it is equally possible to obtain horse in the aeronautical industry. The panel codes wrong results with sophisticated codes as it is possible to require simple surface panels and are not impeded by the obtain smart simulations that help in design decisions requirement of three-dimensional volume grids. from simpler, less-sophisticated codes. Many a times It is important to realize that even when quantitative special features may not leap to the eye from computed accuracy is not guaranteed for all the flow parameters, details of flow, but one needs to closely examine the the CFD codes have still a role to play. For example, it is, detailed flow fields simulated, for example, the pressure even to date difficult to predict drag accurately from any contours, the surface or field velocity vectors, etc. It is

Figure 5. Results for a SARAS wing–body configuration from wind tunnel experiments at NAL in comparison with numerical simu- lations with an Euler and a panel code from the CTFD Division. This kind of comparison justifies the useful role that different levels of approximations of the Navier–Stokes equation play in the preliminary design phases in the aircraft development.

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 53 RESEARCH ACCOUNT then possible to choose from an examination of inviscid ratio. Figure 6 b shows results obtained in a 0.3 m tri- surface pressures, streamline or velocity vector distribu- sonic tunnel at NAL for the pressure distribution over a tions, those that are less prone to flow separation, forma- standard BGK (Bauer, Garabedian and Korn) aerofoil15. tion of strong shocks or favourable pitching moment The shock position was obtained at around 20% of the characteristics, etc. Experience is the key to success in chord for a particular Mach number and incidence such studies. (Figure 6 b). It may be noted here that the accuracies in model making, the dimensional proportions of the model Synergistic use of CFD and wind tunnels to the tunnel height (typically, a ratio of chord to tunnel height less than one-third) and the measurement tech- A synergistic use of CFD and wind tunnels is possible niques were accurate and consistent with recommended and can be gainfully employed in the understanding and test practices for two-dimensional tests. Moreover, the designing of complex hardware such as an aircraft, where results were perfectly repeatable. Because of this, one any one approach may not yield all the answers. It is also could not have suspected their accuracy of simulation of essential to mix the use of many of the novel approximate flow over a given aerofoil. theories of fluid flow. With due experience a synergy of all But then, just around that time it became possible for these can be utilized to bring out a competitive product14. the first time to obtain numerical simulation of viscous It is well known that it is actually the wind tunnel tests transonic flow over this aerofoil by CFD techniques. The carefully carried out that provide the background data for free-air shock location was predicted to be around 65% validation of any numerical simulation software. for the same Mach number and incidence in the numeri- An example is given here, where it was the CFD simu- cal simulation. This created panic and was the basis lations that helped refinements to certain traditional prac- for a CAARC (Commonwealth Advisory Aeronautical tices in wind tunnel testing. Till about the early 80s, it Research Council) programme involving India, Australia, used to be a common practice to use a large open area Canada and England to investigate the role of the open ratio (see Figure 6 a for definition) for the transonic test- area ratio in governing the magnitude of blockage inter- ing of aerofoils within slotted/perforated wall test sec- ference. It was possible to conclude that most of the tun- tions. At NAL we were routinely using 8% open area nels were operating under too open conditions (India 8%,

a b

Figure 6. Pressure distribution for a BGK aerofoil in the 0.3 m tunnel at NAL with 8% open area ratio routinely used till the 80s. Free-air shock position for this aerofoil is around 65% of x/c (see Figure 7 a). a, Schematic diagram of a wind tunnel test section with ventilated walls; b, Experimental results with 8% open area ratio in the 0.3 m trisonic tunnel at NAL. Note that the shock position is at 20% of chord. 8% was routine open area ratio for tests prior to 1980.

54 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT

Australia 20%, England 5%, Canada 12%), leading to surfaces one should have an idea of the control effective- vast tunnel interference of open-jet type. Subsequent ness. For an aft-tail configuration, control effectiveness tests16 at NAL established that an optimum open area of the elevator is of critical importance. This information ratio is indeed around only 2% (see Figure 7 a). As a is not easy to obtain in wind tunnels because of inade- result of this investigation the importance of the role of quate Re simulation; neither is it reliably obtainable from the slots or perforations in the tunnel walls was realized CFD simulations owing to difficulties of geometry (gap for the first time. Indeed, Figure 7 b brings out the dra- effects particularly), turbulence modelling and grid gene- matic effects of the open area ratio on the shock position ration appropriate to the flight Re. It is however possible for a BGK aerofoil, in specific tunnel tests15 which were to obtain fairly easily inviscid simulations, which corre- undertaken to bring out the role of the open area ratio. It spond to a Re of infinity, by using panel methods. Figure 9 is essential to note here that the results with open area adapted from Rubbert17, combines the infinite Re results ratio of 3% and 8% are actually distorted by tunnel with wind tunnel results at two tunnel Re values in a plot boundary conditions and can never be simulated (see Box 1) of control effectiveness versus inverse of Re. From this by CFD with free-air boundary conditions. All the 2D graph it is then possible to read off the control effective- tests are now done only around this reduced open area ness for the required flight Re. ratio16. Figure 8, taken from ref. 15, demonstrates the kind of influence the open area ratio of the tunnel walls Some examples of the use of CFD at NAL18 can have on the shock location over the much-studied NACA 0012 aerofoil at different free stream transonic It is to the credit of the development in CFD that the Mach numbers at zero incidence. software tools developed over the past two decades have This is one example where CFD has actually guided found acceptance in many aircraft development activities and suitably corrected the routine practice in experimen- all over the world. In India, CFD tools are finding increas- tal simulations. ing use in several design and engineering organizations An example is now given where experiments and CFD which include: Hindustan Aeronautics Ltd, Aeronautical are used together to obtain important design data. This Development Agency, Defence Research and Develop- example is taken from Rubbert17. In sizing the control ment Laboratory (DRDL), Naval Science and Techno-

a b

Figure 7. Dramatic role of open area ratio in influencing the aerofoil (BGK) characteristics. Results at 3 and 8% open area ratio are distorted by wall interference and can never be simulated by free-air, far-field conditions in a CFD simulation. a, Experimental results with 2% open area ratio in the 0.3 m trisonic tunnel at NAL. Note the shock position at 65% of chord. Current practice is to use an open area ratio around 2%. b, Dramatic variation in the nature of pressure distributions and in particular shock location as function of open area ratio.

Open air ratio Shock position

8% 20% C 3% 42% C 2% 65% C

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 55 RESEARCH ACCOUNT

Box 1. It is, however, possible to ‘mimic’ these results by suitable modifications to the tunnel wall boundary conditions in the numerical simulations as follows: In a linear fashion one can put the tunnel wall boundary conditions in the most general way as afx + bfy = 0, where x is along the wall and y, normal to the tunnel wall. With this form it is possible to consider the extreme cases, viz. solid wall case with b/a = 0 and open-jet case with a/b = 0, as well as the intermedi- ate cases where a/b ¹ 0. The case a/b ¹ 0 corresponds to slotted or perforated wall conditions. Special tests may be needed to characterize in terms of a and b, a given tunnel wall having slots or perforation with a specified open area ratio. This ‘characterization’ of the tunnel walls is often not easy. In the absence of authentic values for a and b (or for the ratio a/b), values of the ratio of a/b were tried by trial and error in CFD computation to model the wall conditions in the 0.3 m trisonic tunnel at NAL. What is signficant in these simulations is the fact that the results for 2, 3 and 8% open area ratio were closely mimicked by the numerical results. What is still missing is our ability to obtain these values of a/b by any means. The results of the numerical simulations with trial-and-error values of a/b are shown below.

Computed

logical Laboratory, Indian Space Research Organization, CFD simulation, this configuration demonstrated, as in etc. Amongst the academic institutions, mention may be wind tunnel experiments, poor flow aft of the crest of the made of all the Indian Institutes of Technology and the fuselage and over the boom. Subsequently, a CFD study Indian Institute of Science, Bangalore, where CFD is being was undertaken using a panel code12 to finalize a shape used in the analysis of flow fields in research and in that ensured well-organized flow over the aft fuselage sponsored support to other organizations. An overall sce- and boom of the configuration. The modified configura- nario of status of CFD in India is covered in Desai19. tion is the one that went through prototype fabrication, A few examples are presented briefly just to give a fla- flight testing and final type-certification. What is signifi- vour of the latest applications of CFD drawn from cant is that no wind tunnel tests were carried out after the experience at NAL in the aeronautical and the non-aero- modifications suggested by CFD. nautical domain. CFD was again used in the design of wing–body com- bination for the SARAS aircraft20. Figure 11 a shows Aircraft projects computed streamline pattern for the ‘as-designed’ shape. Modifications were carried out to the wing–body inter- The Civil Aircraft projects, viz. HANSA and SARAS, at ference region and to the aft–fuselage near the stub wing NAL have provided opportunities for the CFD to demon- using an Euler code in an analysis mode repeatedly, to strate its utility in aircraft design and analysis. ensure smoothness of streamlines over the configuration. The biggest challenge in aircraft design is to minimize In another representation, Figure 12 b shows the isobars adverse interference effects of one component on the by Euler computations for the two wing–body shapes. As other; obviously, putting together two individually well- can be seen, the isobars smoothly merge into the body designed components may not assure a superior design. isobars in the modified shape. This is evidently the case in the mutual interferences of Figure 13 shows results of Euler computations for the wing and body; independently optimized wing and body wing–fuselage combination pertaining to the MiG-21 may not offer the best wing–body combination. aircraft21. This geometry was created from the production Figure 10 shows the HANSA configuration as put drawings and was analysed for pressure distribution at together by the designer using conventional wisdom. In a transonic Mach numbers to provide load data for studies

56 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT on fatigue life and life-extension to MiG aircraft under- to and beyond stall22. Comparison with wind tunnel data taken at the Structural Integrity Division (SID), NAL. appears to be quite good for lift and pitching moments Figure 14 shows the overall force, drag and moment even up to and beyond stall. It is remarkable, as seen coefficients obtained by the numerical simulation using a from Figure 14, that such a simulation is possible with Re-averaged Navier–Stokes solver for the aerofoil used CFD now. This comparison is further strengthened by the on HANSA aircraft for the entire range of incidences, up comparison of pressure distributions shown in Figure 15.

Figure 9. Procedure for interpolating between wind tunnel and CFD.

a

Modified wing–body shape

Symbol Tunnel Open area c/h Transition

< Slotted, 2% 0.40 Free = NAL 15 × 12” Slotted, 2% Grit 100 at 5% C Baseline 0.40 wing–body 2% wide shape o Slotted, 2% 0.267 Free NAL 15 × 12” Grit 100 at 5% C m Slotted, 2% 0.267 2% wide b D Slotted, 8% 0.40 Free Baseline NAL 15 × 12” Slotted, 8% wing–body ¨ 0.267 Free shape 0.167 TWB Grit 0.07 mm glass Ñ Slotted, 2.35% 0.25 0.6 × 0.34 m 0.33 balls fixed at 7% C Slotted, 9.75% + TWB NPL 5 36 × 14” Slotted, 3.3% 0.278 Grit 230 size CALSPAN ´ Porous 22½% 0.063 Grit at 10% C 96 × 48” Û ARL 31 × 21” Slotted, 16.5% 0.25 Grit at 5%

6 ARL 31 × 21” Slotted, 16.5% 0.125 Grit at 5% Figure 10. Use of panel code in HANSA design. a, Use of a panel code in designing wing–body junction for proper flow management of

the rear fuselage. b, Computed surface pressure contours for HANSA Figure 8. Shock location from various tunnels (NACA 0012, a = 0°). aircraft at NAL.

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 57 RESEARCH ACCOUNT

What was most essential from an application point of model design and fabrication, availability and timeliness view in one instance was information on the movement of tunnel test schedules normally persuade investigators of the front stagnation point as a function of the aero- to seek CFD solutions as an alternative. foil/wing incidence. Figure 16 shows the data obtained Figure 19 shows the numerical simulations from a low- for the HANSA aerofoil. This information, useful in the speed RANS code25 for the tethered balloon under a location of a stall warning system, cannot be easily sponsorship. The pressure distributions and the subse- obtained in a wind tunnel simulation. quent load information obtained by CFD were the direct Figure 17 shows the results of an effort in which CFD inputs to the structural analysis groups at NAL. was used to reduce the areas of pressure drag of SARAS In another application, information on pressure distri- aircraft by local geometric changes1. It is believed that bution and loads was generated for a 13-m diameter this exercise is valuable and could bring about a reduc- weather radome26 that was to be designed and built by tion of up to 5% in pressure drag. But such local changes NAL for a sponsor. Figure 20 shows the pressure distri- are impossible to be studied expeditiously in a wind butions obtained on the spherical radome mounted on a tunnel simulation. tall building. The simulations not only provided informa- In the hypersonic flow regime where the wind tunnel tion about overall load and wind pressure, but also about simulations begin to become impractical, design help will the level of truncation of the spherical radome to improve come increasingly from the CFD simulations. Figure 18 the lift-to-drag ratio for the radome. This increased lift- gives the density contours at different sections of air inlet to-drag ratio helped in the structural design of the of a Hypersonic Research Vehicle23,24. radome. What is noteworthy in this effort has been that no wind tunnel tests were conducted nor contemplated. Non-aircraft applications

The present-day scenario in project execution is marked by a tendency of the investigators to seek alternatives to tunnel testing. Time and cost of accurate wind-tunnel

Figure 11. Streamline patterns from Euler computations: wing–body (SARAS) (a) as designed, (b) after modifications suggested by CFD. Figure 12. Isobars from Euler computations: wing–body (SARAS). Notice the smooth pattern of streamlines on the fuselage around and Notice the smooth merger of wing isobars with the isobars over the downstream of the wing region. fuselage around the wing region.

58 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT

Figure 13. Aerodynamic load data for life extension (at SID, NAL).

That is the kind of confidence one can place on CFD simulations; of course, this level depends on the geomet- ric complexity of the shapes and also on the speed regimes of flow.

Future trends in wind tunnel testing

Figure 21, adapted from open literature, brings together the tunnel testing requirements for some of the well- known commercial airliners. For example, typically for the development of Boeing 747, wind tunnel hours of the order of about 10,000 were required during the late six- ties and the early seventies. The trend in Figure 21 shows a continuous increase in the demand for wind tunnel test- ing up to mid-eighties. In Figure 21 we have indicated

Figure 14. Aerodynamic characteristics for GA(W)-2 aerofoil. ¾, N–S computation; ¡¡¡, WSU tests; DDD, NASA tests. What is re- t markable is these comparisons is that they hold for lift and pitching moment up to and beyond stall; drag is still an evasive parameter in CFD simulation.

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 59 RESEARCH ACCOUNT

Figure 15. Surface pressure distribution for GA(W)-2 aerofoil ¾, N–S computation; OOO, WSU tests; DDD, NASA tests. These good com- parisons for the pressure distributions are actually behind the good comparisons for pitching moment up to and beyond stall shown in Figure 17. Levels of pressure drag on SARAS aircraft. Figure 13.

Figure 16. Stagnation point location for GA(W)-2 aerofoil (M¥ = 0.3, Re = 2 × 106). This information may not be obtained easily in wind Figure 18. RANS computations at Mach 6.5 – density contours at tunnel tests. different sections of air inlet of a Hypersonic Research Vehicle.

60 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT the possible reduction in the wind tunnel test require- experience can help in bringing about a drastic reduction ments due to CFD support subsequent to 1985. It is per- in the requirements for tunnel testing. haps not unreasonable to realize that even with full It is useful to go back to Figure 4 where it is mentioned support from CFD capabilities the number of hours of that a typical aircraft design would entail something like tunnel testing may not come down below about 5000 h, 2.5 million simulations. It is also mentioned that some 10 for a new design. Of course, for a derivative design, past to 100 CFD simulations might add considerable value to

c d

· Multiblock RANS computation on parallel computer using 300 × 82 × 88 (approximately 2 million) nodes with H–O grid topology. · Realistic surface pressure distribution and close agreement between measurement data and RANS prediction are observed for a 20°, = 0°, Roll = 0°.

Figure 19. Turbulent flow around an aerostat with fins (Re = 1.5 × 107).

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 61 RESEARCH ACCOUNT

· O–O Grid topology (162 × 52 × 62) used (Reynolds No. = 4.789 × 107). · Realistic surface pressure distribution obtained. · Flow separation on dome surface followed by a typical longitudinal vortex pair in the wake cap- tured reasonably well. · Within an error band the present computation gives (CL) consistent to those from other sources including wind tunnel data.

Figure 20. Turbulent flow around a radome structure.

62 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 RESEARCH ACCOUNT

Figure 21. Wind tunnel testing requirements for aircraft development programmes and future support from CFD.

Table 1. Some estimates of simulation costs for CFD and wind tunnel tests at NAL

Simulation Tool Time for one simulation Cost per++ simulation (Rs)

Panel code; symmetric case; 5000 panels Half an hour on O-200* 1000 Asymmetric case; 10,000 panels Half an hour on FLOSOLVER** 2500 CFD Euler code (wing–body 1m grids) 8 h on O-200 16,000 RANS (wing–body only) 20 h on O-200 40,000 1.2 m tunnel*** at NAL Wind tunnel – 1500+ (only overall forces and moments)

*Computer charges at Rs 2000/h. **FLOSOLVER charges at Rs 5000/h. ***Tunnel charges at Rs 15,000/blow-down; model making cost not included. +Assume ten simulations in one blow-down. ++Cost here does not include geometry and grid generation costs in CFD and model making costs in wind tunnel testing. these large number of simulations. After a wind tunnel a synergy from CFD, to arrive at an optimum design model is made, tunnel simulations are fast and less ex- sooner and cheaper than in the earlier times. pensive compared to each of the CFD simulations. It is virtually impossible to hope that several hundred simula- 1. Desai, S. S., Curr. Sci., 1999, 77, 1283–1294. tions could practically be carried out by CFD in an 2. Rubbert, Paul, E., AIAA Lecture, Anaheim, affordable time-frame for an aircraft design. This is evi- California, 18–23 September 1994. 3. Desai, S. S., Rangarajan, R., Sinha, U. N. and Ravichandran, dent from Table 1, where some rough estimates are given K. S., Symposium Transsonicum III, May 1988, pp. 109–120. for the cost of CFD and wind tunnel simulations at NAL. 4. Rangarajan, R. and Desai, S. S., Proceedings of the NAL-DFVLR Workshop on Design Aerodynamics, NAL SP 8608, May 1986. Concluding remarks 5. Srilatha, K. R., Dwarakanth, G. S. and Ramamoorthy, P., J. Air- craft, 1990, 27, 1.2.1–1.2.9. The laudable growth over the past three decades in the 6. Srilatha, K. R. and Ramamoorthy, P., Report, Design of a Laminar capability of CFD has convincingly brought home its Flow for an Unmanned Aircraft, NAL PD CF-9004, 1990. distinctive role as an adjuvant to the design and deve- 7. Nair, M. T. and Saxena, S. K., Report, Upwind Implicit Residual lopment efforts in aerospace. Its role in aeronautical Smoothing for a Multi-block Upwind and Reynolds Averaged Navier-Stokes Solver, NAL PD CF 9913, Sept. 1999. engineering is to be judged by its addition of efficacy to 8. Chakrabatty, S. K., Dhanalakshmi, K. and Mathur, J. S., Acta the traditional elements of aircraft design, such as engi- Mech., 1998, 131, 69–87. neering data sheets and experiences, wind tunnel testing 9. Dutta, Vimala, Proceedings of the First Asian Computational and flight testing. It is not to be expected that these tradi- Fluid Dynamics Conference, Hong-Kong, 16–17 January 1995, tional methods will be put to disuse because of the vol. 2, pp. 683–690. 10. Dutta, Vimala, Gopinath, R. and Dutta, P. K., Proceedings of CFD mature developments in computational fluid dynamics; it 97, University of Victoria, British Columbia, Canada, 25–27 May is indeed important to realize that these traditional meth- 1997. ods will continue to have their roles to play but now with 11. Dutta Vimala, J. Aero. Soc. India, 1999, 51, 238–251.

CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003 63 RESEARCH ACCOUNT

12. Viswanathan, S., Narayana, C. L., Ramesh, V., Desai, S. S. and 22. Dutta, Vimala and Desai, S. S., Report, Position for stall warning Rangarajan, R., Proceedings of the Fluid Dynamics Symposium in system on HANSA-3 Prototype II, NAL PD CF 9905, May honour of R. Narasimha on his 60th birthday, NAL SP 9315, July 1999. 1993, pp. 211–233. 23. Nair, Manoj, T., Naresh Kumar and Saxena, S. K., Report, Rey- 13. Chakrabartty, S. K., Dhanalakshmi, K. and Mathur, J. S., Proceed- nolds Averaged Navier–Stokes based Aerodynamic Analysis of a ings Third Asian CFD Conference, 7–11 December 1998, vol. 2, Hypersonic Research Vehicle, NAL PD CF 0207, June 2002. pp. 126–131. 24. Nair, M. T. and Saxena, S. K., Proceedings 8th International 14. Desai, S. S., Invited Talk 4th Asian Fluid Dynamics Conference, Workshop on Shock Tube Technology, Indian Institute of Science, Mianyang, China, 18–22 September 2000. Bangalore, 11–14 September 2002. 15. Desai, S. S., Ramachandra Sharma, Venkateswarlu, K. and Krish- 25. Majumdar, S., Report, CFD Analysis of Viscous Flow Around namurthy, K., Report, Wall Interference Studies on NACA 0012 Aerostat Configuration using Pressure-based RANS Code, NAL Aerofoil in the NAL 0. 3m Wind Tunnel, NAL-TM-5–82, 1982. PD CF 9902, February 1999. 16. Sundara Murthy, H., Desai, S. S. and Suryanarayana, G. K., Paper 26. Majumdar, S. and Rajani, B. N., Proceedings 4th International 3.2, NAL-DFVLR Workshop of Design Aerodynamics, NAL, Symposium on Engineering Turbulence Modelling and Measure- Bangalore, 7–9 May 1986. ments, Corsica, France, 24–26 May 1999. 17. Rubbert, P., Transsonium III, IUTAM Symposium, Göttingen, 1988. 18. Desai, S. S., Report, Recent Progress at NAL, PD CF 9807, November 1998. ACKNOWLEDGEMENTS. It is a pleasure to acknowledge the many 19. Desai, S. S., Computational Fluid Dynamics, Status Report in technical discussions I have had with my colleagues at NAL: Dr S. K. Some Selected Topics in Theoretical and Applied Mechanics (ed. Chakrabartty, Dr M. D. Deshpande, Dr S. Majumdar, Dr J. S. Mathur, Nakra, B. C.), Diamond Jubilee Publication, Indian National Sci- Dr V. Y. Mudkavi, Dr P. N. Shankar, Dr S. K. Saxena from the CTFD ence Academy, 1994, pp. 1–35. Division; Dr P. R. Viswanath from Experimental Aerodynamics Divi- 20. Chakrabartty, S. K., Dhanalakshmi, K. and Mathur, J. S., Curr. son; Dr M. Shivakumara Swamy, Dr K. Yegna Narayan from Centre Sci., 1999, 77, 1295–1302. for Civil Aircraft Design and Development. Thanks are due to V. S. 21. Mathur, J. S., Dhanalakshmi, K. and Chakrabartty, S. K., Report, Narasimhan and Mallikarjuna for help in preparing this manuscript. Euler Analysis of MiG-21 Wing Fuselage Configurations, PD CF 9907, July 1999. Received 11 June 2002; revised accepted 21 November 2002

RESEARCH ARTICLE Mycobacterium leprae 18-kDa heat shock protein gene is polymorphic A. K. Shabaana*, N. P. Shankernarayan# and K. Dharmalingam*,† *Department of Genetic Engineering, School of Biotechnology, Madurai Kamaraj University, Madurai 625 021, India #Voluntary Health Services, Leprosy Project, Shakti Nagar 638 315, India

Leprosy is still a major health problem in India. Myco- TCA in about 60% of the samples and as CCA in rest bacterium leprae, the organism causing the disease, of the leprosy cases. Armadillo-derived M. leprae 18- appears to be genetically invariant. Attempts to iden- kDa HSP gene has TCA at the 52nd position. Among tify strain variation in different isolates of M. leprae the cases examined, we could not detect mixed popula- have met with little success. The M. leprae 18-kDa tion of bacteria carrying both classes of genes, indicat- heat shock protein is a major T-cell antigen, and this ing the clonal purity of the bacterium in an infected protein belongs to the family of small heat shock pro- patient. This polymorphism therefore, could be used teins. In the present study, we have analysed the as a molecular marker in strain typing of M. leprae in mRNA expression profiles of 18-kDa gene in lesions endemic populations. In addition, transmission pat- across the leprosy spectra by RT–PCR and also se- terns and relapse and/or reinfection in reactional quenced the PCR amplicons prepared from 25 inde- patients could be examined using this polymorphism. pendent leprosy cases. A single nucleotide poly- Expression of the 18-kDa HSP gene in reactional cases morphism was detected at 154th bp position in this indicates that live bacterium might be contributing to secreted antigen gene. In this gene, codon 52 exists as the reactional conditions in leprosy.

1 LEPROSY is a chronic infectious disease caused by the bal leprosy burden . Leprosy is an immunological disease acid-fast bacillus Mycobacterium leprae. Leprosy is still with defined immunological and clinical parameters, and a health problem globally, and India has 64% of the glo- ranges from tuberculoid (TT) to lepromatous (LL) lep- rosy with a range of borderline cases between the polar ends2. TT patients have robust, cell-mediated immune †For correspondence. (e-mail: [email protected]) response that restricts the growth of the pathogen leading

64 CURRENT SCIENCE, VOL. 84, NO. 1, 10 JANUARY 2003